Knowledge Graph Construction for Intelligent Analysis of Social Networking User Opinion

Microblogging is a popular social networking tool on which people tend to express their views and opinions. As such, the massive data on microblogging platforms mean abundant research value to social science researchers. To help them better analyze these data, a framework for understanding diverse user opinions and identifying complex relationships in the form of knowledge graphs is proposed in this paper. The two main tasks in the framework are sentiment analysis and knowledge graph construction. In the first task, the Skip-gram model is employed to obtain the word embedding matrix and the Bi-LSTM model is adopted to perform stance classification. It is found in this paper that Bi-LSTM showed better performance in classifying different sentiments, compared with Naive Bayes and SnowNLP. In the second task, relations between different users are extracted from their micro-blogs through recognizing specific strings, and on this basis user attitudes are integrated into the knowledge extracted. A knowledge graph of user opinions is constructed with the Neo4J tool. With the knowledge extracted by this framework, social science researchers can more easily observe rules of perspective communication and perform further analysis of the data.